A wireless sensor system for a biofeedback training of hammer throwers
© The Author(s) 2016
Received: 8 April 2016
Accepted: 12 August 2016
Published: 22 August 2016
Hammer-throw has a long-standing history in track and field, but unlike some other sports events, men’s hammer throw has not seen a new world record since 1986. One of the possible reasons for this stagnation could be the lack of real-time biomechanical feedback training. In this study, we proposed to establish scientifically described training targets and routes, which in turn required tools that could measure and quantify characteristics of an effective hammer-throw. Towards this goal, we have developed a real-time biomechanical feedback device—a wireless sensor system—to help the training of hammer-throw. The system includes two sensors—an infrared proximity sensor for tracing the hip vertical movement and a load cell for recording the wire tension during a hammer-throw. The system uses XBees for data transmission and an Arduino processor for data processing and system control. The results revealed that the wire tension measurement could supply sufficient key features for coaches to analyze hammer-throw and give real-time feedback for improving training efficiency.
Effective human motor skill learning/training not only benefits athletes but can also promote more active lifestyles in the general population (Chen and Ennis 2004; Li et al. 2016; Wan and Shan 2016). The two key components in motor learning are practice and biofeedback (Schmidt and Lee 2011). Previous studies have shown that, when properly understood and applied, biofeedback can strongly enhance the practice of human motor skills (Shan et al. 2004; Visentin et al. 2008). Generally, there are three types of biofeedback: physiological (e.g. heart rate), neurological (e.g. EEG/brain-wave), and biomechanical (e.g. joint angles and force applied) (Tate and Milner 2010). While physiological and neurological feedback devices are commonly seen in practice, biomechanical feedback devices are still in their developing phase. The reasons for the current situation could be the following points: (1) effective biomechanical feedback should relate to the invisible forces (i.e. we can only feel the effect of a force, but cannot see it; the only way for its visualization/quantification is through a force measurement device, such as a scale) controlling the limb movement of human motor skills (Shan and Westerhoff 2005), and (2) motor skill learning and optimization must be tailored to an individual body structure and the activity being examined (Shan and Bohn 2003; Shan et al. 2015; Visentin et al. 2015). Because of the advance of wearable/wireless sensor systems, a “tailored” biomechanical feedback device would be theoretically possible for individualized training. Such a device would allow the users to self-correct problems based on such biofeedback (such as the invisible forces during a movement) provided by feedback devices.
One of the possible tools for biomechanical feedback training is wireless sensor networks (WSNs). WSNs are composed of distributed nodes connected with sensors which communicate with each other and send/receive data to the base station. Each sensor node has a battery for power, a microprocessor for programming, and a transceiver for communication. In this paper, the proposed architecture is based on Xbee/Arduino modules where Xbee is used for communication and Arduino used to control and process the data. WSNs have been applied in a wide range of applications, such as in agriculture (Keshtgari and Deljoo 2012), in health care monitoring (Mansor et al. 2013), in smart home technology (Lu et al. 2014), in environment observation (Lazarescu 2012) and ecosystem (Du et al. 2015). The successes of WSNs in the above areas suggest that their application potential in human motor learning and training is high.
One of the practical challenges for establishing the biomechanical feedback device is its size, not only the sensor size but also the microprocessor unit. It should be tiny and wearable, but at the same time, it should not constrain an athlete’s movement. IEEE 802.15.4-compliant (Digi 2014a) transceivers are typically used for communications in WSNs. The performance of smart grid applications with IEEE 802.15.4-compliant transceivers has been studied by Bilgin and Gungor (2012). Piyare and Lee (2013) analyzed the efficiency of XBee ZB module-based WSNs regarding the received signal strength indication (RSSI), delay, throughput, energy, etc.
Several successful application examples using the above technology have been reported. They range from collecting climatologic data (Keshtgari and Deljoo 2012), measuring body temperature development and heart rate in patients (Mansor et al. 2013; Kioumars and Tang 2011), and monitoring radioactive materials (Ding et al. 2009). These applications share the communication protocols (ZigBee) and RF hardware, but not necessarily the underlying computing platform.
In these examples, the sensors are relatively static, and on the receiver side, the computer is connected to another XBee module to make the wireless communication available, and the computer is used to monitor and process the data. However, for a biofeedback application in human movement, further development is required.
As discussed above, biomechanical feedback must be tailored to a specific activity. In the current study, the hammer throw is chosen for the development of the biomechanical feedback device. Hammer throw has a long-standing history in track and field, but unlike some other events, hammer throw has not seen a new world record since 1986 (IAAF 2015). One of the possible reasons for this stagnation could be the lack of scientific feedback data needed for the training. While extensive 3D motion analysis technologies do exist for hammer throw, practitioners have reported that they are too cumbersome to be useful for training (Shan et al. 2012). The main issue is the time-consuming data collection and processing. Such a procedure would make biomechanical feedback available for practitioners after weeks, reducing the practicality of 3D motion caption in hammer throw training. Due to the complexity of the movement and the difficulty of its data collection, hardly any scientific research exists for the hammer throw. Therefore, a real-time biofeedback tool, e.g. a wireless wearable sensor system, is desired in practice.
Therefore, the current study aimed to develop a new, wearable real-time biomechanical feedback device, which would measure (1) real-time wire tension and (2) vertical hip displacement. Specifically, the hardware development aimed to prototype a wireless data collection unit and the software development intended to equip the feedback device with software that can collect wire tension and vertical hip displacement measurements, receive and store tension and displacement data, perform primary data processing functions and include a graphical user interface for real-time data visualization. Such a feedback system would have further potentials for development to (1) establish how to reach desirable tension and displacement during a throw, and ultimately (2) provide biomechanically-guided training plans customized to each athlete’s anthropometrical data. In short, the system developed would have great potential to be both a research tool for better understanding of hammer throw movements and a user-friendly training tool for coaches and athletes.
Results and discussions
The aims of the study were to develop a wearable biomechanical feedback device. We have successfully prototyped the device. The applicability of the device was tested in the training sessions of the Canadian hammer throw team. The Human Subjects Research Committee of the University of Lethbridge/Canada scrutinized and approved the protocols as meeting the criteria of ethical conduct for research involving humans. The athletes were informed that the device would be used for collecting data related to their throws. They signed an approved consent form and voluntarily participated in the data collection.
The male data is from the Canadian Champion’s throw. His best performance, achieved in May 2008 in Lethbridge, Alberta, still stands as the current Canadian record. In Fig. 2, one can see that there are several peaks before reaching the maximum release point. The video capture showed that the throw could be divided into two phases: (1) preparation—the subject pulled up the hammer from the ground and then, swung the hammer for two circles before starting the next phase; (2) body turning—the subject performed four and a half turns before the release of the hammer. The first three peaks represented the preparation phase—Peak 1: pull-up, Peak 2 and 3: two circles of hammer swing. The preparation ended at 2.8 s, and then the athlete entered the turning phase.
The video data unveiled that the turning consisted of double-support and single-support. Double-support is the duration of each turn where both feet are in contact with the ground. Conversely, single-support is the portion of each turn where the right foot (for a right-handed thrower) is in the air while the left foot remains in contact with the ground. Based on physics, a thrower needs to extend double-support as much as possible because the only way to increase speed is to drive or push with the right foot and the right foot is only on the ground during the double-support, so this is truly the only time to accelerate body rotation. By lengthening the double-support time (while shortening the single-support time), one can push effectively, resulting in increasing the angular velocity of the body-turning thus increasing the speed of the ball. Therefore, it can increase the length of the flight.
The results of rotary speed in each turn and the release speed from the Canadian champion’s field tests
Wire tension (N)
Max rotary speed in each turn (m/s)
Increase of rotary speed in each turn (%)
Turn 1 (Peak 4)
Turn 2 (Peak 5)
Turn 3 (Peak 6)
Turn 4 (Peak 7)
The release (Max)
The results of rotary speed in each turn and the release speed from a female subject’s field tests
Wire tension (N)
Max rotary speed in each turn (m/s)
Increase of rotary speed in each turn (%)
Turn 1 (Peak 4)
Turn 2 (Peak 5)
Turn 3 (Peak 6)
Turn 4 (Peak 7)
The release (Max)
Which control strategy would be more effective? What is the optimized control for hammer throw? Should males and females use a different throw technique? Should the control pattern be individualized according to one’s physical condition? The results suggest more questions than supply solutions for the questions. Definitely, more applied studies using biofeedback device in training are needed to answer the questions. However, the current study implies that coaches could use the real time, biomechanical feedback tool to experiment various/possible motor control strategies for skill optimization in practice.
Inevitably, there are flaws in the prototyping. The current optical distance sensor requires that the sensor points vertically towards the ground to get the correct distance, i.e. the up-and-down movement of the hip. During the field tests, it was found that hip-orientation changed continuously during the turns; as such, the wireless distance sensor could not supply valid data. Further studies using alternative distance sensors are needed to investigate the hip movement to remedy the flaw for adding hip control into feedback in learning and motor skill optimization. Additionally, it is planned to use Bluetooth technology to implement the receiver node in a cell phone or a mobile device, such as iPhone, iPad or other tablets. In this case, coaches will have a more convenient way to perform the real-time biofeedback training. Finally, an improvement on the current structure or designing a PCB (Printed Circuit Board) is planned to minimize the device for more convenience to the athletes.
Wireless sensor networks have great application potential in human motor skill learning and optimization. Using the example of the hammer throw, we have shown in the current study that properly designed WSN device could supply invisible control information of professional athletes. Such valuable information would help coaches establish real-time biofeedback training and improve the performance of athletes. Most importantly, the current study extends the WSNs application to a new area—the professional athletes training.
Hardware and system configuration
The basic idea in our research is to establish a system of the WSN to receive two kinds of data: the distance from the athlete’s waist (hip) to the ground—up and down movement of the upper body—and the tension during the process of the hammer throw. In this system, we use a sensor node to collect data and send the data to a receiver node via wireless communication.
Figure 6 shows the hardware. It contains three basic components: two sensors, an Arduino board, and an XBee. There are two sensors in this system. One is for measuring the distance from the waist of the athlete to the ground. We use the infrared proximity sensor made by Sharp, which has an analog output varying from 2.8 V at 15 cm to 0.4 V at 150 cm (Sparkfun 2014). The other one is a load cell (tension sensor) for testing the wire tension during hammer throw. The load cell is produced by Omegadyne, and we use the type of LCFD-1 K, which can measure as high as 5000 N (Omegadyne 2014).
In our design, the data sending actions are event-triggered in order to reduce the cost of data communication. Once the tension sensor’s analog signals reach above 15 units (that is about 20 N), which means the athlete starts a throw, the sensor device will be able to start working and collect data automatically. By using the event-trigger, we resolved the issues of how to avoid receiving junk data.
Another critical idea used in our sensor node is an easy-release connector installed between the hammer and our system device. When the hammer is thrown away by the athlete, the connector will be released along with the hammer, which means the tension sensor cannot feel any tension at this time. Then the sensor node will stop collecting/transmitting data. During the athlete performs a movement, the sensor node keeps sending data to the receiver node in real-time. By applying the two important techniques, our system will be convenient and accurate for both coaches and athletes.
The receiver node is shown in Fig. 8. The receiver node consists of an XBee, which is the same type of the one used in the sensor node, and a laptop. The XBee in the receiver node also needs a shield, which is right under the XBee module (Fig. 9), to be connected to the computer via a USB cable. The end-user computer is used to receive, monitor and process the data sent from the sensor node.
Programming and interface
In the sensor node, we can write codes in Arduino sketch which is a kind of software integrated development environment based on C/C++. Its library is related with AVR Libc and allows people to use its functions (Arduino 2014b). We can upload the program directly from the Arduino sketch to our Arduino Mega board so that we can control our sensor node and make an initial process when collecting and sending data. Our program was implemented based on the AnalogReadSerial (Arduino 2014c).
In the receiver node, we use MATLAB as our programming tool. We create a graphical user interface to show the data values of the two sensors in MATLAB. We also implement a program to process the data and plot the data. For minimizing electrical noise, the Butterworth filter is used when plotting the data. After the release (i.e. the sensor node was separated from the microcontroller), the collected data is sent to the receiver node, and it is filtered for a real-time plot in MATLAB. The cut-off frequency of the Butterworth filter in MATLAB is set to 0.2, which can provide a smooth and reasonable curve. The MATLAB GUI program is used to monitor and process real-time data on PC.
YW designed and programmed the sensor system and tested the performance. BW helped the field tests, analyzed and interpreted the data. HL and GS proposed the architecture and improved the design. GS secured the research funding. All authors participated in the manuscript writing. All authors read and approved the final manuscript.
Ye Wang is currently a Ph.D. student of computer science at the University of Lethbridge. His research topic is about wireless sensor network and its application. He finished his undergraduate study at the University of Lethbridge in 2012 and obtained his master degree at the same University in 2015. He was a co-op student in the Research Centre of Agriculture and Agri-Food Canada from January 2013 to April 2013, working on a project related to WSNs application in agriculture. Bingjun Wan obtained his Ph.D. in coaching science from Beijing Sports University, Beijing, China. He is now an Associate Professor in the School of Physical Education, Shaanxi Normal University, Xi’an, China. Dr. Wan’s research projects are related to training efficiency and injury prevention during motor learning and training. His research projects are supported by the General Administration of Sport of China. Currently, Dr. Wan is a China Scholarship Council (CSC) exchange scholar at the University of Lethbridge/Canada and doing collaborative research with Prof. Dr. Shan. Hua Li has been working at the University of Lethbridge after obtained Ph.D. in Computer Science. He is an Associate Professor in the department of mathematics and computer science. His current research interests include wireless sensor networks, VLSI and hardware design, FPGA, and network security. Gongbing Shan obtained his Ph.D. in biomechanics from the University of Münster, Germany. After his Ph.D., he was an NIH (National Institute of Health/USA) Post-Doctoral Fellow in the University of Vermont/USA. He is now a Professor (Biomechanics) at the University of Lethbridge, Alberta, Canada. Dr. Shan is the founder of a state-of-the-art Biomechanics Lab at the university. The lab is equipped with a high-speed 3D motion capture system, wireless EMG, force platforms and bicycle pedal force sensors along with other equipment personally developed by Dr. Shan. Since its establishment in 2000, the lab has been hosting a wide range of high-impact, interdisciplinary research projects. The projects are supported by Canada Foundation for Innovation (CFI), Natural Sciences and Engineering Research Council of Canada (NSERC), Covenant Health Research Center (CHRC) and Southern Alberta Intellectual Property Network (SAIPN).
This work was supported by National Sciences and Engineering Research Council of Canada (NSERC). We would like to thank the Canadian Hammer Throw Team for applying our system device into their training sessions and help with the field tests.
The authors declare that they have no competing interests.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
- Arduino (2014a) Arduino Mega. http://arduino.cc/en/Main/arduinoBoardMega. Accessed 16 June 2014
- Arduino (2014b) Language reference. http://arduino.cc/en/Reference/HomePage. Accessed 16 June 2014
- Arduino (2014c) Analog read serial. http://arduino.cc/en/Tutorial/AnalogReadSerial. Accessed 16 June 2014
- Bilgin B, Gungor V (2012) Performance evaluations of ZigBee in different smart grid environments. Comput Netw 56:2196–2205View ArticleGoogle Scholar
- Chen A, Ennis CD (2004) Goals, interests, and learning in physical education. J Educ Res 97(6):329–338View ArticleGoogle Scholar
- Digi (2014a) XBee/RF solutions. http://www.digi.com/products/xbee-rf-solutions. Accessed 7 Mar 2014
- Digi (2014b) XCTU: next generation configuration platform for XBee/RF solutions. http://www.digi.com/products/wireless-wired-embedded-solutions/zigbee-rf-modules/xctu. Accessed 7 Mar 2014
- Ding F, Song G, Yin K, Li J, Song A (2009) A GPS-enabled wireless sensor network for monitoring radioactive materials. Sens Actuat A Phys 155:210–215View ArticleGoogle Scholar
- Du W, Xing Z, Li M, He B, Chua L, Miao H (2015) Sensor placement and measurement of wind for water quality studies in urban reservoirs. ACM Trans Sens Netw 11:167–178View ArticleGoogle Scholar
- IAAF (2015) Hammer throw—all time best. http://www.iaaf.org/records/toplists/throws/hammer-throw/outdoor/men/senior. Accessed 26 Mar 2015
- Keshtgari M, Deljoo A (2012) A wireless sensor network solution for precision agriculture based on zigbee technology. Lect Notes Comput Sci 4:25–30Google Scholar
- Kioumars A, Tang L (2011) Wireless network for health monitoring: heart rate and temperature sensor. Sens Technol (ICST). doi:10.1109/ICSensT.2011.6137000 Google Scholar
- Lazarescu M (2012) Design of a WSN platform for long-term environmental monitoring for IoT applications. IEEE J Emerg Sel Top Circuits Syst 3:45–54View ArticleGoogle Scholar
- Li S, Zhang Z, Wan B, Wilde B, Shan G (2016) The relevance of body positioning and its training effect on badminton smash. J Sports Sci. doi:10.1080/02640414.2016.1164332 Google Scholar
- Lu J, Shams Y, Whitehouse K (2014) Smart blueprints: how simple sensors can collaboratively map out their own locations in the home. ACM T Sens Netw 11:19.1–19.23Google Scholar
- Mansor H, Shukor M, Meskam S, Rusli N, Zamery N (2013) Body temperature measurement for remote health monitoring system. IEEE ICSIMA. doi:10.1109/ICSIMA.2013.6717956 Google Scholar
- Omegadyne (2014) Subminiature tension or compression high accuracy load cells. http://www.omegadyne.com/pdf/lcfd.pdf. Accessed 26 Jan 2014
- Piyare R, Lee S (2013) Performance analysis of XBee ZB module based wireless sensor networks. Int J Sci Eng Res 4:1615–1621Google Scholar
- Schmidt R, Lee T (2011) Motor control and learning: a behavioral emphasis, 5th edn. Human Kinetics, ChampaignGoogle Scholar
- Shan G, Bohn C (2003) Anthropometrical data and coefficients of regression related to gender and race. Appl Erg 34:327–337View ArticleGoogle Scholar
- Shan G, Westerhoff P (2005) Full-body kinematic characteristics of the maximal instep soccer kick by male soccer players and parameters related to kick quality. Sports Biomech 4:59–72View ArticleGoogle Scholar
- Shan G, Sust M, Simard S, Bohn C, Nicol K (2004) How can dynamic rigid-body modeling be helpful in motor learning? Diagnosing performance using dynamic modeling. Kinesiology 36:182–191Google Scholar
- Shan G, Yuan J, Hao W, Gu M, Zhang X (2012) Regression equations related to the quality evaluation of soccer maximal instep kick for males and females. Kinesiology 44:139–147Google Scholar
- Shan G, Visentin P, Zhang X, Hao W, Yu D (2015) Bicycle kick in soccer: is the virtuosity systematically entrainable? Sci Bull 60:819–821View ArticleGoogle Scholar
- Sparkfun (2014) Product description. https://www.sparkfun.com/products/8958. Accessed 6 Jan 2014
- Tate J, Milner C (2010) Real-time kinematic, temporospatial, and kinetic biofeedback during gait retraining in patients: a systematic review. Phys Ther 90:1123–1134View ArticleGoogle Scholar
- Visentin P, Shan G, Wasiak EB (2008) Informing music teaching and learning using movement analysis technology. Int J Music Educ 26:73–87View ArticleGoogle Scholar
- Visentin P, Li S, Tardif G, Shan G (2015) Unraveling mysteries of personal performance style; biomechanics of left-hand position changes (shifting) in violin performance. PeerJ 3:e1299View ArticleGoogle Scholar
- Wan B, Shan G (2016) Biomechanical modeling as a practical tool for predicting injury risk related to repetitive muscle lengthening during learning and training of human complex motor skills. SpringerPlus. doi:10.1186/s40064-016-2067-y Google Scholar